area-science-park/ESM3-PPISites

Protein-Protein Interaction Site Prediction

This model is a finetuned version of ESM3-open for protein-protein interaction site prediction. It predicts whether a certain amino acid in a protein sequence is part of an interaction site (1) or not (0).

• 📄 For more details on the training and testing on this model, refer to the pre-print here: https://doi.org/10.64898/2026.05.29.728739.
• 📦 The github repository to use with this model is available here: https://github.com/RitAreaSciencePark/ESM3-PPISites
• 📊 The data for the training and evaluation of this model is available in csv format in this zenodo repository: https://doi.org/10.5281/zenodo.20366948

How to Get Started with the Model

Important note: you need to have access to the esm3-open model (https://huggingface.co/biohub/esm3-sm-open-v1) to use this model.

import torch
from transformers import AutoModel, AutoTokenizer, AutoConfig
from esm.tokenization import get_esm3_model_tokenizers

model_name = "area-science-park/ESM3-PPISites"
model = AutoModel.from_pretrained(model_name, trust_remote_code=True)

tokenizers = get_esm3_model_tokenizers("esm3_sm_open_v1")
sequence_tokenizer = tokenizers.sequence

# move model to device
device = "cuda" if torch.cuda.is_available() else "cpu"
model = model.to(device)

# run over a sample sequence
sequence = "MKTVRQERLKSIVRILEAAKEPVSGAQLAEELSVSRQVIVQDIAYLRSLGYNIVATPRGYVLAGG"
# tokenize
tokens = sequence_tokenizer.encode(sequence)
inputs = torch.tensor(tokens).unsqueeze(0).to(device)
# inference
logits = model(inputs)["logits"]
probabilities = torch.sigmoid(logits)

# get predictions
probabilities

References

Decoding the Grammar of Protein-Protein Interaction Interfaces with Multimodal Representations. Yuri Gardinazzi*, Edith Natalia Villegas Garcia*, Sergio Senci, Davide Di Vora, Antonio Feltrin, Francesca Cuturello. bioRxiv 2026.05.29.728739; doi: https://doi.org/10.64898/2026.05.29.728739

Built with ESM3. License: Cambrian Non-Commercial License Agreement (https://www.evolutionaryscale.ai/policies/cambrian-non-commercial-license-agreement)